小白努力学编程

Netty5源码分析--1.服务端启动过程详解

13 January 2014

实例

样例代码来自于io.netty.example.telnet.TelnetServer，完整样例请参考NettyExample工程。

public class TelnetServer {

private final int port;

public TelnetServer(int port) {
    this.port = port;
}

public void run() throws Exception {
    EventLoopGroup bossGroup = new NioEventLoopGroup();//bossGroup线程池用来接受客户端的连接请求
    EventLoopGroup workerGroup = new NioEventLoopGroup();//workerGroup线程池用来处理boss线程池里面的连接的数据
    try {
        ServerBootstrap b = new ServerBootstrap();
        b.group(bossGroup, workerGroup)
         .channel(NioServerSocketChannel.class)
         .childHandler(new TelnetServerInitializer());//ChannelInitializer是一个特殊的handler，用来初始化ChannelPipeline里面的handler链。 这个特殊的ChannelInitializer在加入到pipeline后，在initChannel调用结束后,自身会被remove掉，从而完成初始化的效果（后文会详述）。

//AbstractBootstrap.option()用来设置ServerSocket的参数，AbstractBootstrap.childOption()用来设置Socket的参数。

        b.bind(port).sync().channel().closeFuture().sync();
    } finally {
        bossGroup.shutdownGracefully();
        workerGroup.shutdownGracefully();
    }
}

public static void main(String[] args) throws Exception {
    int port;
    if (args.length > 0) {
        port = Integer.parseInt(args[0]);
    } else {
        port = 8080;
    }
    new TelnetServer(port).run();
}
}

针对上述代码，还需要补充介绍一些内容：

在调用ctx.write(Object)后需要调用ctx.flush()方法，这样才能将数据发出去。或者直接调用 ctx.writeAndFlush(msg)方法。

通常使用这种方式来实例化ByteBuf：final ByteBuf time = ctx.alloc().buffer(4); ，而不是直接使用ByteBuf子类的构造方法

另外，还需要在处理基于流的传输协议TCP/IP的数据时，注意报文和业务程序实际能够接收到的数据之间的关系。 假如你发送了2个报文，底层是发送了两组字节。但是操作系统的TCP栈是有缓存的，它可能把这两组字节合并成一组字节，然后再给业务程序使用。但是业务程序往往需要根据把这一组字节还原成原来的两组字节，但是不幸的是，业务程序往往无法直接还原，除非在报文上做了些特殊的约定。比如报文是定长的或者有明确的分隔符。

服务端启动服务

当`TelnetServer启动时，依次完成如下步骤：

NioEventLoopGroup初始化

当NioEventLoopGroup构造方法被调用时，首先初始化父类MultithreadEventLoopGroup,触发父类获得默认的线程数，其值默认是Runtime.getRuntime().availableProcessors() * 2

static {
    DEFAULT_EVENT_LOOP_THREADS = Math.max(1, SystemPropertyUtil.getInt(
            "io.netty.eventLoopThreads", Runtime.getRuntime().availableProcessors() * 2));

    if (logger.isDebugEnabled()) {
        logger.debug("-Dio.netty.eventLoopThreads: {}", DEFAULT_EVENT_LOOP_THREADS);
    }
}

接着调用NioEventLoopGroup自身的构造器，依次执行下面的构造器。

public NioEventLoopGroup() {
    this(0);
}

public NioEventLoopGroup(int nThreads, Executor executor) {
    this(nThreads, executor, SelectorProvider.provider());
}

public NioEventLoopGroup(int nThreads) {
    this(nThreads, (Executor) null);
}

public NioEventLoopGroup(int nThreads, ThreadFactory threadFactory) {
    this(nThreads, threadFactory, SelectorProvider.provider());
}

public NioEventLoopGroup(
        int nThreads, ThreadFactory threadFactory, final SelectorProvider selectorProvider) {
    super(nThreads, threadFactory, selectorProvider);
}

继续调用父类MultithreadEventLoopGroup的构造器，该构造器又调用了父类构造器。

    protected MultithreadEventLoopGroup(int nThreads, Executor executor, Object... args)    {
    super(nThreads == 0 ? DEFAULT_EVENT_LOOP_THREADS : nThreads, executor, args);
}

下面的构造方法主要完成以下几件事情：

  1. 设置默认DefaultThreadFactory线程工厂，主要做了2件事，设置线程池名称和线程名称

  2. 初始化children数组，然后通过调用NioEventLoopGroup.newChild方法完成child属性设置。

    protected MultithreadEventExecutorGroup(int nThreads, Executor executor, Object... args) {
if (nThreads <= 0) {
    throw new IllegalArgumentException(String.format("nThreads: %d (expected: > 0)", nThreads));
}

if (executor == null) {
    executor = new ThreadPerTaskExecutor(newDefaultThreadFactory());
}

children = new EventExecutor[nThreads];
for (int i = 0; i < nThreads; i ++) {
    boolean success = false;
    try {
        children[i] = newChild(executor, args);//tag
        success = true;
    } catch (Exception e) {
        // TODO: Think about if this is a good exception type
        throw new IllegalStateException("failed to create a child event loop", e);
    } finally {
        if (!success) {
            for (int j = 0; j < i; j ++) {
                children[j].shutdownGracefully();
            }

            for (int j = 0; j < i; j ++) {
                EventExecutor e = children[j];
                try {
                    while (!e.isTerminated()) {
                        e.awaitTermination(Integer.MAX_VALUE, TimeUnit.SECONDS);
                    }
                } catch (InterruptedException interrupted) {
                    Thread.currentThread().interrupt();
                    break;
                }
            }
        }
    }
}

    在newChild方法中，主要完成构建NioEventLoop实例

    @Override
protected EventLoop newChild(Executor executor, Object... args) throws  Exception {
return new NioEventLoop(this, executor, (SelectorProvider) args[0]);
}

    下面的super(parent, executor, false);主要是设置NioEventLoopGroup是NioEventLoop的parent。然后调用openSelector()创建Selector对象。

    NioEventLoop(NioEventLoopGroup parent, Executor executor, SelectorProvider selectorProvider) {
super(parent, executor, false);
if (selectorProvider == null) {
    throw new NullPointerException("selectorProvider");
}
provider = selectorProvider;
selector = openSelector();
}

首先，先初始化Selector对象，然后再初始化SelectedSelectionKeySet，设置其属性keysA = new SelectionKey[1024]; keysB = keysA.clone();。进行了一个优化，设置了sun.nio.ch.SelectorImpl的selectedKeys和publicSelectedKeys属性。根据NioEventLoop.run()方法内部直接调用 processSelectedKeysOptimized(selectedKeys.flip());并且没有直接使用selector.selectedKeys()这两处代码，笔者猜测正是因为在此时通过反射设置了属性，所以NioEventLoop.run()才能正常工作。

  private Selector NioEventLoop.openSelector() {
    final Selector selector;
    try {
        selector = provider.openSelector();
    } catch (IOException e) {
        throw new ChannelException("failed to open a new selector", e);
    }

    if (DISABLE_KEYSET_OPTIMIZATION) {
        return selector;
    }

    try {
        SelectedSelectionKeySet selectedKeySet = new SelectedSelectionKeySet();

        Class<?> selectorImplClass =
                Class.forName("sun.nio.ch.SelectorImpl", false, ClassLoader.getSystemClassLoader());

        // Ensure the current selector implementation is what we can instrument.
        if (!selectorImplClass.isAssignableFrom(selector.getClass())) {
            return selector;
        }

        Field selectedKeysField = selectorImplClass.getDeclaredField("selectedKeys");
        Field publicSelectedKeysField = selectorImplClass.getDeclaredField("publicSelectedKeys");

        selectedKeysField.setAccessible(true);
        publicSelectedKeysField.setAccessible(true);

        selectedKeysField.set(selector, selectedKeySet);
        publicSelectedKeysField.set(selector, selectedKeySet);

        selectedKeys = selectedKeySet;
        logger.trace("Instrumented an optimized java.util.Set into: {}", selector);
    } catch (Throwable t) {
        selectedKeys = null;
        logger.trace("Failed to instrument an optimized java.util.Set into: {}", selector, t);
    }

    return selector;
}

最后循环完成children数组的初始化children[i] = newChild(executor, args);，进而完成NioEventLoopGroup对象初始化。

小结

此时再结合Eclipse的DEBUG视图，观察bossGroup的属性，可以基本看到完成如下几个事情

  • 创建NioEventLoopGroup对象
  • 获得默认线程池数目大小，数值为N
  • 设置线程池名称和线程名称
  • 循环创建出来 N个NioEventLoop对象，每个NioEventLoop都设置了相同的parent，executor和不同的selector实例。

ServerBootstrap 初始化

ServerBootstrap b = new ServerBootstrap();

上面这段代码内涵平平，主要设置group属性是bossGroup，childGroup属性是workerGroup。 没啥其他复杂属性赋值。主要值得一提的就是channel方法的设计，通过传递class对象，然后通过反射来实例化具体的Channel实例。

b .bind(port) .sync() .channel() .closeFuture() .sync();，这个方法的内容很多，详见下述分析。

b.bind(port)方法会调用下面的doBind方法，在doBind方法中会完成Channel的初始化和绑定端口。有2个方法需要tag，分别是 tag1 和 tag2

private ChannelFuture doBind(final SocketAddress localAddress) {
    final ChannelFuture regFuture = initAndRegister();//tag1
    final Channel channel = regFuture.channel();
    if (regFuture.cause() != null) {
        return regFuture;
    }

    final ChannelPromise promise;
    if (regFuture.isDone()) {
        promise = channel.newPromise();
        doBind0(regFuture, channel, localAddress, promise);//tag2
    } else {
        // Registration future is almost always fulfilled already, but just in case it's not.
        promise = new DefaultChannelPromise(channel, GlobalEventExecutor.INSTANCE);
        regFuture.addListener(new ChannelFutureListener() {
            @Override
            public void operationComplete(ChannelFuture future) throws Exception {
                doBind0(regFuture, channel, localAddress, promise);
            }
        });
    }

    return promise;
}

tag1 initAndRegister，里面完成Channel实例创建，实例化和注册channel到selector上。

final ChannelFuture initAndRegister() {
    Channel channel;
    try {
        channel = createChannel();//tag1.1
    } catch (Throwable t) {
        return VoidChannel.INSTANCE.newFailedFuture(t);
    }

    try {
        init(channel);//tag1.2
    } catch (Throwable t) {
        channel.unsafe().closeForcibly();
        return channel.newFailedFuture(t);
    }

    ChannelPromise regFuture = channel.newPromise();
    channel.unsafe().register(regFuture);//tag1.3
    if (regFuture.cause() != null) {
        if (channel.isRegistered()) {
            channel.close();
        } else {
            channel.unsafe().closeForcibly();
        }
    }

tag1.1,调用ServerBootstrap.createChannel() ，通过反射完成Channel实例创建。这里使用了childGroup这个属性，即workGroup线程池。

@Override
Channel createChannel() {
    EventLoop eventLoop = group().next();
    return channelFactory().newChannel(eventLoop, childGroup);//tag1.1.1

}

tag1.1.1，此时将断点打到NioServerSocketChannel的构造方法上

 public NioServerSocketChannel(EventLoop eventLoop, EventLoopGroup childGroup) {
    super(null, eventLoop, childGroup, newSocket(), SelectionKey.OP_ACCEPT);//tag1.1.1.1
    config = new DefaultServerSocketChannelConfig(this, javaChannel().socket());//tag1.1.1.2

}

tag1.1.1.1,这段代码主要完成3件事。

第一个是在NioServerSocketChannel.newSocket()调用了ServerSocketChannel.open()，完成了javaChannel的创建

    private static ServerSocketChannel newSocket() {
    try {
        return ServerSocketChannel.open();
    } catch (IOException e) {
        throw new ChannelException(
                "Failed to open a server socket.", e);
    }
}

第二个是在AbstractNioChannel的构造方法中调用了ch.configureBlocking(false)方法

 protected AbstractNioChannel(Channel parent, EventLoop eventLoop, SelectableChannel ch, int readInterestOp) {
    super(parent, eventLoop);//tag1.1.1.1.1
    this.ch = ch;
    this.readInterestOp = readInterestOp;
    try {
        ch.configureBlocking(false);
    } catch (IOException e) {
        try {
            ch.close();
        } catch (IOException e2) {
            if (logger.isWarnEnabled()) {
                logger.warn(
                        "Failed to close a partially initialized socket.", e2);
            }
        }

        throw new ChannelException("Failed to enter non-blocking mode.", e);
    }
}

tag1.1.1.1.1中，在AbstractChannel(Channel parent, EventLoop eventLoop)中，进行了两个重要操作：unsafe = newUnsafe();pipeline = new DefaultChannelPipeline(this);。

protected AbstractChannel(Channel parent, EventLoop eventLoop) {
    this.parent = parent;
    this.eventLoop = validate(eventLoop);
    unsafe = newUnsafe();
    pipeline = new DefaultChannelPipeline(this);//tag1.1.1.1.1.1
}

tag1.1.1.1.1.1,设置了HeadHandler和TailHandler。这两个类也比较重要。

public DefaultChannelPipeline(AbstractChannel channel) { if (channel == null) { throw new NullPointerException("channel"); } this.channel = channel;

    TailHandler tailHandler = new TailHandler();
    tail = new DefaultChannelHandlerContext(this, null, generateName(tailHandler), tailHandler);//tag1.1.1.1.1.1.1

    HeadHandler headHandler = new HeadHandler(channel.unsafe());
    head = new DefaultChannelHandlerContext(this, null, generateName(headHandler), headHandler);

    head.next = tail;
    tail.prev = head;
}

tag1.1.1.1.1.1.1,这个方法完成了DefaultChannelHandlerContext的对象的初始化。这个类也是核心类，先暂时把它当成个黑盒，会在后面重点分析。

此时，我们方法调用栈结束，然后回到 tag1.1.1.2 这段代码上来。 在DefaultServerSocketChannelConfig中构造方法中完成了channel的参数设置

至此，才完成tag1.1 AbstractBootstrap.createChannel()方法的执行。现在又开始 tag1.2的代码片段。该 AbstractBootstrap.init(Channel channel) 方法里面主要涉及到Parent Channel 和 Child Channel的option和attribute 设置，并将客户端设置的参数覆盖到默认参数中；最后，还将childHandler(new TelnetServerInitializer())中设置的handler加入到pipeline()中。代码见下。

void init(Channel channel) throws Exception { final Map<ChannelOption<?>, Object> options = options(); synchronized (options) { channel.config().setOptions(options); }

    final Map<AttributeKey<?>, Object> attrs = attrs();
    synchronized (attrs) {
        for (Entry<AttributeKey<?>, Object> e: attrs.entrySet()) {
            @SuppressWarnings("unchecked")
            AttributeKey<Object> key = (AttributeKey<Object>) e.getKey();
            channel.attr(key).set(e.getValue());
        }
    }

    ChannelPipeline p = channel.pipeline();
    if (handler() != null) {
        p.addLast(handler());
    }

    final ChannelHandler currentChildHandler = childHandler;
    final Entry<ChannelOption<?>, Object>[] currentChildOptions;
    final Entry<AttributeKey<?>, Object>[] currentChildAttrs;
    synchronized (childOptions) {
        currentChildOptions = childOptions.entrySet().toArray(newOptionArray(childOptions.size()));
    }
    synchronized (childAttrs) {
        currentChildAttrs = childAttrs.entrySet().toArray(newAttrArray(childAttrs.size()));
    }

    p.addLast(new ChannelInitializer<Channel>() {//tag1.2.1
        @Override
        public void initChannel(Channel ch) throws Exception {
            ch.pipeline().addLast(new ServerBootstrapAcceptor(currentChildHandler, currentChildOptions,
                    currentChildAttrs));
        }
    });
}

tag1.2.1中，此时pipeline中又多了一个handler：内部类ServerBootstrap$1，此时数组的链表情况如下：HeadHandler，ServerBootstrap$1和TailHandler。另外，再额外吐槽一句，p.addLast方法并不是把ServerBootstrap$1放到tail上，而是放到tail的前一个节点上。所以，这个addLast方法命名很是误解。

至此完成tag1.2执行，开始执行tag1.3 channel.unsafe().register(regFuture);这段代码。该方法内部接着执行执行tag1.3.1的代码。

public final void register(final ChannelPromise promise) {
        if (eventLoop.inEventLoop()) {
            register0(promise);
        } else {
            try {
                eventLoop.execute(new Runnable() {
                    @Override
                    public void run() {//tag1.3.1
                        register0(promise);
                    }
                });
            } catch (Throwable t) {
                logger.warn(
                        "Force-closing a channel whose registration task was not accepted by an event loop: {}",
                        AbstractChannel.this, t);
                closeForcibly();
                closeFuture.setClosed();
                promise.setFailure(t);
            }
        }
    }

tag1.3.1,该片段主要执行doRegister();和pipeline.fireChannelRegistered();//tag1.3.1.2

private void register0(ChannelPromise promise) {
        try {
            // check if the channel is still open as it could be closed in the mean time when the register
            // call was outside of the eventLoop
            if (!ensureOpen(promise)) {
                return;
            }
            doRegister();//tag1.3.1.1
            registered = true;
            promise.setSuccess();
            pipeline.fireChannelRegistered();//tag1.3.1.2
            if (isActive()) {
                pipeline.fireChannelActive();
            }
        } catch (Throwable t) {
            // Close the channel directly to avoid FD leak.
            closeForcibly();
            closeFuture.setClosed();
            if (!promise.tryFailure(t)) {
                logger.warn(
                        "Tried to fail the registration promise, but it is complete already. " +
                                "Swallowing the cause of the registration failure:", t);
            }
        }
    }

tag1.3.1.1 将代码片段将javachannel注册到selector上，并把selectionKey属性赋值

protected void AbstractNioChannel.doRegister() throws Exception {
    boolean selected = false;
    for (;;) {
        try {
            selectionKey = javaChannel().register(eventLoop().selector, 0, this);
            return;
        } catch (CancelledKeyException e) {
            if (!selected) {
                // Force the Selector to select now as the "canceled" SelectionKey may still be
                // cached and not removed because no Select.select(..) operation was called yet.
                eventLoop().selectNow();
                selected = true;
            } else {
                // We forced a select operation on the selector before but the SelectionKey is still cached
                // for whatever reason. JDK bug ?
                throw e;
            }
        }
    }
}

tag1.3.1.2，这个方法里面有一堆事情要讲。先暂且放下，在后文讲到ChannelPipeline时会再次回来看这段代码。

public ChannelPipeline DefaultChannelPipeline.fireChannelRegistered() { head.fireChannelRegistered(); return this; }

此时终于完成 tag1 代码片段执行，开始执行 tag2 的代码片段。

private static void doBind0( final ChannelFuture regFuture, final Channel channel, final SocketAddress localAddress, final ChannelPromise promise) {

    // This method is invoked before channelRegistered() is triggered.  Give user handlers a chance to set up
    // the pipeline in its channelRegistered() implementation.
    channel.eventLoop().execute(new Runnable() {
        @Override
        public void run() {
            if (regFuture.isSuccess()) {
                channel.bind(localAddress, promise).addListener(ChannelFutureListener.CLOSE_ON_FAILURE);//tag2.1
            } else {
                promise.setFailure(regFuture.cause());
            }
        }
    });
}

tag2.1，该方法内部有调用了pipeline的方法了（在tag1.3.1.2 中也出现了pipeline调用）。 好吧，是时候介绍pipeline了。

public ChannelFuture bind(SocketAddress localAddress, ChannelPromise promise) { return pipeline.bind(localAddress, promise);//tag2.1.1 }

ChannelPipeline

DefaultChannelPipeline是ChannelPipeline的实现类，DefaultChannelPipeline内部维护了两个指针：final DefaultChannelHandlerContext head; final DefaultChannelHandlerContext tail;，分别指向链表的头部和尾部；而DefaultChannelHandlerContext内部是一个链表结构：volatile DefaultChannelHandlerContext next;volatile DefaultChannelHandlerContext prev;，而每个DefaultChannelHandlerContext与ChannelHandler实例一一对应。

从上面可以看到，这是个经典的Intercepting Filter模式实现。下面我们再接着从tag1.3.1.2代码看起，pipeline.fireChannelRegistered();依次执行如下两个方法。上文也已经说明，此时handler链是HeadHandler，ServerBootstrap$1和TailHandler。

   @Override
public ChannelPipeline DefaultChannelPipeline.fireChannelRegistered() {
    head.fireChannelRegistered();
    return this;
}

public ChannelHandlerContext ChannelHandlerContext.fireChannelRegistered() {
    DefaultChannelHandlerContext next = findContextInbound(MASK_CHANNEL_REGISTERED); //tag 1.3.1.2.1
    next.invoker.invokeChannelRegistered(next); //tag1.3.1.2.2

    return this;
}

private DefaultChannelHandlerContext DefaultChannelHandlerContext.findContextInbound(int mask) {
    DefaultChannelHandlerContext ctx = this;
    do {
        ctx = ctx.next;
    } while ((ctx.skipFlags & mask) != 0);
    return ctx;
}

tag 1.3.1.2.1,针对这个findContextInbound方法需要再补充下，里面ServerBootstrap$1是继承自ChannelInitializer，而ChannelInitializer.channelRegistered是没有@Skip注解的。呃，@Skip注解又有何用。这个要结合DefaultChannelHandlerContext.skipFlags0(Class<? extends ChannelHandler> handlerType)。这个skipFlags0方法返回一个整数，如果该方法上标记了@Skip注解，那么表示该方法在Handler被执行时，需要被忽略。所以，此时do {ctx = ctx.next;} while ((ctx.skipFlags & mask) != 0);片段的执行结果返回的是ServerBootstrap$1这个Handler。

这里在额外说一句，这个ChannelHandlerAdapter里面的方法几乎都被加了@Skip标签。

private static int skipFlags0(Class<? extends ChannelHandler> handlerType) {
    int flags = 0;
    try {
        if (handlerType.getMethod(
                "handlerAdded", ChannelHandlerContext.class).isAnnotationPresent(Skip.class)) {
            flags |= MASK_HANDLER_ADDED;
        }
        if (handlerType.getMethod(
                "handlerRemoved", ChannelHandlerContext.class).isAnnotationPresent(Skip.class)) {
            flags |= MASK_HANDLER_REMOVED;
        }
        if (handlerType.getMethod(
                "exceptionCaught", ChannelHandlerContext.class, Throwable.class).isAnnotationPresent(Skip.class)) {
            flags |= MASK_EXCEPTION_CAUGHT;
        }
        if (handlerType.getMethod(
                "channelRegistered", ChannelHandlerContext.class).isAnnotationPresent(Skip.class)) {
            flags |= MASK_CHANNEL_REGISTERED;
        }
        if (handlerType.getMethod(
                "channelActive", ChannelHandlerContext.class).isAnnotationPresent(Skip.class)) {
            flags |= MASK_CHANNEL_ACTIVE;
        }
        if (handlerType.getMethod(
                "channelInactive", ChannelHandlerContext.class).isAnnotationPresent(Skip.class)) {
            flags |= MASK_CHANNEL_INACTIVE;
        }
        if (handlerType.getMethod(
                "channelRead", ChannelHandlerContext.class, Object.class).isAnnotationPresent(Skip.class)) {
            flags |= MASK_CHANNEL_READ;
        }
        if (handlerType.getMethod(
                "channelReadComplete", ChannelHandlerContext.class).isAnnotationPresent(Skip.class)) {
            flags |= MASK_CHANNEL_READ_COMPLETE;
        }
        if (handlerType.getMethod(
                "channelWritabilityChanged", ChannelHandlerContext.class).isAnnotationPresent(Skip.class)) {
            flags |= MASK_CHANNEL_WRITABILITY_CHANGED;
        }
        if (handlerType.getMethod(
                "userEventTriggered", ChannelHandlerContext.class, Object.class).isAnnotationPresent(Skip.class)) {
            flags |= MASK_USER_EVENT_TRIGGERED;
        }
        if (handlerType.getMethod(
                "bind", ChannelHandlerContext.class,
                SocketAddress.class, ChannelPromise.class).isAnnotationPresent(Skip.class)) {
            flags |= MASK_BIND;
        }
        if (handlerType.getMethod(
                "connect", ChannelHandlerContext.class, SocketAddress.class, SocketAddress.class,
                ChannelPromise.class).isAnnotationPresent(Skip.class)) {
            flags |= MASK_CONNECT;
        }
        if (handlerType.getMethod(
                "disconnect", ChannelHandlerContext.class, ChannelPromise.class).isAnnotationPresent(Skip.class)) {
            flags |= MASK_DISCONNECT;
        }
        if (handlerType.getMethod(
                "close", ChannelHandlerContext.class, ChannelPromise.class).isAnnotationPresent(Skip.class)) {
            flags |= MASK_CLOSE;
        }
        if (handlerType.getMethod(
                "read", ChannelHandlerContext.class).isAnnotationPresent(Skip.class)) {
            flags |= MASK_READ;
        }
        if (handlerType.getMethod(
                "write", ChannelHandlerContext.class,
                Object.class, ChannelPromise.class).isAnnotationPresent(Skip.class)) {
            flags |= MASK_WRITE;

            // flush() is skipped only when write() is also skipped to avoid the situation where
            // flush() is handled by the event loop before write() in staged execution.
            if (handlerType.getMethod(
                    "flush", ChannelHandlerContext.class).isAnnotationPresent(Skip.class)) {
                flags |= MASK_FLUSH;
            }
        }
    } catch (Exception e) {
        // Should never reach here.
        PlatformDependent.throwException(e);
    }

    return flags;
}

此时，tag1.3.1.2.1 代码片段执行完毕，现在开始tag1.3.1.2.2 执行。

@Override
public void DefaultChannelHandlerInvoker.invokeChannelRegistered(final ChannelHandlerContext ctx) {
    if (executor.inEventLoop()) {
        invokeChannelRegisteredNow(ctx);
    } else {
        executor.execute(new Runnable() {
            @Override
            public void run() {
                invokeChannelRegisteredNow(ctx);
            }
        });
    }
}

public static void invokeChannelRegisteredNow(ChannelHandlerContext ctx) {
    try {
        ctx.handler().channelRegistered(ctx);
    } catch (Throwable t) {
        notifyHandlerException(ctx, t);
    }
}

由于ServerBootstrap$1(ChannelInitializer<C>)这个类继承了ChannelInitializer，所以会执行了ChannelInitializer.channelRegistered这个方法。

@Override
@SuppressWarnings("unchecked")
public final void ChannelInitializer.channelRegistered(ChannelHandlerContext ctx) throws Exception {
    ChannelPipeline pipeline = ctx.pipeline();
    boolean success = false;
    try {
        initChannel((C) ctx.channel());//tag1.3.1.2.2.1
        pipeline.remove(this);//tag1.3.1.2.2.2
        ctx.fireChannelRegistered();//tag1.3.1.2.2.3
        success = true;
    } catch (Throwable t) {
        logger.warn("Failed to initialize a channel. Closing: " + ctx.channel(), t);
    } finally {
        if (pipeline.context(this) != null) {
            pipeline.remove(this);
        }
        if (!success) {
            ctx.close();
        }
    }
}

在tag1.3.1.2.2.1里，又回调了下面的initChannel方法。该方法把ServerBootstrapAcceptor这个Handler加入到Pipeline中；此时handler链情况如下：HeadHandler，ServerBootstrap$1，ServerBootstrap$ServerBootstrapAcceptor和TailHandler

 p.addLast(new ChannelInitializer<Channel>() {
        @Override
        public void initChannel(Channel ch) throws Exception {
            ch.pipeline().addLast(new ServerBootstrapAcceptor(currentChildHandler, currentChildOptions,
                    currentChildAttrs));
        }
    });

在 tag1.3.1.2.2.2里，通过执行pipeline.remove(this);又把ServerBootstrap$1这个Handler给删除了，从而完成初始化的效果。需要提醒的是，ServerBootstrapAcceptor的currentChildHandler属性包含了在客户端代码注册的TelnetServerInitializer类。

在tag1.3.1.2.2.3里，通过执行ctx.fireChannelRegistered();又找到了下一个handler，

public ChannelHandlerContext DefaultChannelHandlerContext.fireChannelRegistered() { DefaultChannelHandlerContext next = findContextInbound(MASK_CHANNEL_REGISTERED); next.invoker.invokeChannelRegistered(next); return this; }

这段逻辑和上述基本一样， findContextInbound内部执行时，会跳过ServerBootstrapAcceptor这个handler，最终找到找到tailHandler，并执行channelRegistered()这个方法。就这样，最终完成了整个 pipeline.fireChannelRegistered();执行。

static final class TailHandler extends ChannelHandlerAdapter {

@Override
public void channelRegistered(ChannelHandlerContext ctx) throws Exception {}

//省略下面的方法

}

下面我们再趁热打铁，回头看看 tag2.1代码的执行逻辑。

 public ChannelFuture AbstractChannel.bind(SocketAddress localAddress, ChannelPromise promise)  {
    return pipeline.bind(localAddress, promise);//tag2.1.1
} 


 @Override
public ChannelFuture DefaultChannelPipeline.bind(SocketAddress localAddress, ChannelPromise promise) {
    return pipeline.bind(localAddress, promise);
}

 @Override
public ChannelFuture bind(SocketAddress localAddress, ChannelPromise promise) {
    return tail.bind(localAddress, promise); //tag2.1.1.1
}

tag2.1.1.1，执行到这里，发现是tail.bind，而不是head.bind。

 @Override
public ChannelFuture bind(final SocketAddress localAddress, final ChannelPromise promise) {
    DefaultChannelHandlerContext next = findContextOutbound(MASK_BIND);
    next.invoker.invokeBind(next, localAddress, promise);
    return promise;
}

@Override public void DefaultChannelHandlerInvokerinvokeBind( final ChannelHandlerContext ctx, final SocketAddress localAddress, final ChannelPromise promise) { if (localAddress == null) { throw new NullPointerException("localAddress"); } validatePromise(ctx, promise, false);

    if (executor.inEventLoop()) {
        invokeBindNow(ctx, localAddress, promise);
    } else {
        safeExecuteOutbound(new Runnable() {
            @Override
            public void run() {
                invokeBindNow(ctx, localAddress, promise);
            }
        }, promise);
    }
}


 public static void ChannelHandlerInvokerUtil.invokeBindNow(
        final ChannelHandlerContext ctx, final SocketAddress localAddress, final ChannelPromise promise) {
    try {
        ctx.handler().bind(ctx, localAddress, promise);
    } catch (Throwable t) {
        notifyOutboundHandlerException(t, promise);
    }
}

@Override
public void DefaultChannelPipeline.HeadHandler.bind(
            ChannelHandlerContext ctx, SocketAddress localAddress, ChannelPromise promise)
            throws Exception {
        unsafe.bind(localAddress, promise);
    }

 @Override
 public final void AbstractChannel.AbstractUnsafe.bind(final SocketAddress localAddress, final ChannelPromise promise) {
        if (!ensureOpen(promise)) {
            return;
        }

        // See: https://github.com/netty/netty/issues/576
        if (!PlatformDependent.isWindows() && !PlatformDependent.isRoot() &&
            Boolean.TRUE.equals(config().getOption(ChannelOption.SO_BROADCAST)) &&
            localAddress instanceof InetSocketAddress &&
            !((InetSocketAddress) localAddress).getAddress().isAnyLocalAddress()) {
            // Warn a user about the fact that a non-root user can't receive a
            // broadcast packet on *nix if the socket is bound on non-wildcard address.
            logger.warn(
                    "A non-root user can't receive a broadcast packet if the socket " +
                    "is not bound to a wildcard address; binding to a non-wildcard " +
                    "address (" + localAddress + ") anyway as requested.");
        }

        boolean wasActive = isActive();
        try {
            doBind(localAddress);//tag2.1.1.1.1
         } catch (Throwable t) {
            promise.setFailure(t);
            closeIfClosed();
            return;
        }
        if (!wasActive && isActive()) {
            invokeLater(new Runnable() {//tag2.1.1.1.2
                @Override
                public void run() {
                    pipeline.fireChannelActive();//tag2.1.1.1.3
                }
            });
        }
        promise.setSuccess();//tag2.1.1.1.4
    }

在tag2.1.1.1.1里，执行真正的bind端口。

protected void doBind(SocketAddress localAddress) throws Exception {
    javaChannel().socket().bind(localAddress, config.getBacklog());
}    

在tag2.1.1.1.2里，执行如下方法，`eventLoop().execute(task); `在后续分析。现在暂时忽略。  
private void invokeLater(Runnable task) {
        // This method is used by outbound operation implementations to trigger an inbound event later.
        // They do not trigger an inbound event immediately because an outbound operation might have been
        // triggered by another inbound event handler method.  If fired immediately, the call stack
        // will look like this for example:
        //
        //   handlerA.inboundBufferUpdated() - (1) an inbound handler method closes a connection.
        //   -> handlerA.ctx.close()
        //      -> channel.unsafe.close()
        //         -> handlerA.channelInactive() - (2) another inbound handler method called while in (1) yet
        //
        // which means the execution of two inbound handler methods of the same handler overlap undesirably.
        eventLoop().execute(task);
    }

这里需要说一下，虽然先执行了invokeLater该方法，但是仅仅是把给task加入到队列中，然后等 tag2.1.1.1.4 方法执行后，在下一个循环中再继续执行。

@Override
public ChannelPromise DefaultChannelPromise.setSuccess() {
    return setSuccess(null);
}

@Override
public ChannelPromise setSuccess(Void result) {
    super.setSuccess(result);
    return this;
}

@Override
public Promise<V> setSuccess(V result) {
    if (setSuccess0(result)) {// tag2.1.1.1.4.1
        notifyListeners();// tag2.1.1.1.4.2
        return this;
    }
    throw new IllegalStateException("complete already: " + this);
}

 private boolean setSuccess0(V result) {
    if (isDone()) {
        return false;
    }

    synchronized (this) {
        // Allow only once.
        if (isDone()) {
            return false;
        }
        if (result == null) {
            this.result = SUCCESS;// tag2.1.1.1.4.1.1
        } else {
            this.result = result;
        }
        if (hasWaiters()) {
            notifyAll();
        }
    }
    return true;
}

在 tag2.1.1.1.4.1.1 设置了成功状态，然后该方法返回，继续执行了tag2.1.1.1.4.2方法。由于listeners为 null，所以直接返回。

private void notifyListeners() {

    Object listeners = this.listeners;
    if (listeners == null) {
        return;
    }
    // 省略XXXXXX
}

此时，程序完成了tag2.1 代码执行，开始继续循环。此时执行 tag2.1.1.1.3里的代码，即执行pipeline.fireChannelActive();方法。

  public ChannelPipeline fireChannelActive() {
    head.fireChannelActive();//tag2.1.1.1.3.1

    if (channel.config().isAutoRead()) {
        channel.read();//tag2.1.1.1.3.2
    }

    return this;
}

在tag2.1.1.1.3.1里，和上述逻辑一样，最终执行到TailHandler这里。

static final class TailHandler extends ChannelHandlerAdapter {

    @Override
    public void channelRegistered(ChannelHandlerContext ctx) throws Exception { }

    @Override
    public void channelActive(ChannelHandlerContext ctx) throws Exception { }

    //下省略方法

}

在tag2.1.1.1.3.2里，由于channel.config().isAutoRead()默认返回true；

 @Override
public ChannelPipeline read() {
    tail.read();
    return this;
}

 @Override
    public void DefaultChannelPipeline.HeadHandler.read(ChannelHandlerContext ctx) {
        unsafe.beginRead();
    }


 @Override
 public void AbstractChannel.AbstractUnsafe.beginRead() {
        if (!isActive()) {
            return;
        }

        try {
            doBeginRead();
        } catch (final Exception e) {
            invokeLater(new Runnable() {
                @Override
                public void run() {
                    pipeline.fireExceptionCaught(e);
                }
            });
            close(voidPromise());
        }
    }

此属性 readInterestOp值为16，interestOps & readInterestOp值为0，所以执行了selectionKey.interestOps(interestOps | readInterestOp);，等同于执行了selectionKey.interestOps(SelectionKey.OP_ACCEPT);。

 protected void AbstractNioChannel.doBeginRead() throws Exception {
    if (inputShutdown) {
        return;
    }

    final SelectionKey selectionKey = this.selectionKey;
    if (!selectionKey.isValid()) {
        return;
    }

    final int interestOps = selectionKey.interestOps();
    if ((interestOps & readInterestOp) == 0) {
        selectionKey.interestOps(interestOps | readInterestOp);
    }
}

至此，整个DefaultPromise.bind方法执行完毕，下面开始执行DefaultPromise.sync()。而此时在 tag2.1.1.1.4.1.1 已经将值设为SUCCESS了，所以不需要等待，直接返回。

 @Override
public Promise<V> DefaultPromise.sync() throws InterruptedException {
    await();
    rethrowIfFailed();
    return this;
}

 @Override
public Promise<V> DefaultPromise.await() throws InterruptedException {
    if (isDone()) {
        return this;
    }

    if (Thread.interrupted()) {
        throw new InterruptedException(toString());
    }

    synchronized (this) {
        while (!isDone()) {
            checkDeadLock();
            incWaiters();
            try {
                wait();
            } finally {
                decWaiters();
            }
        }
    }
    return this;
}

然后系统接着执行了 b.bind(port).sync().channel().closeFuture().sync();的后半截方法“channel().closeFuture().sync()”方法。而由于closeFuture这个属性的执行结果一直没有赋值，所以被wait了，从而一直处于wait状态。

至此，主线程处于wait状态，并通过子线程无限循环，来完成客户端请求。

小结

通过channel方法设置不同的通道类型，通过childHandler设置SocketChannel的Handler链

bind(port)完成的职责很多，远不同于ServerSocket.bind方法。具体包含：initAndRegister和doBind0。

其中initAndRegister又细化了createChannel() 和init(channel)以及channel.unsafe().register(regFuture)这3个大步骤。

  • createChannel内部 使用了childGroup，group().next()，ServerSocketChannel.open()这3个属性来创建NioServerSocketChannel实例，并初始化了默认参数DefaultServerSocketChannelConfig和DefaultChannelPipeline对象。DefaultChannelPipeline对象默认包含设置了HeadHandler和TailHandler。然后设置了ch.configureBlocking(false)模式，并将readInterestOp赋值为SelectionKey.OP_ACCEPT。

  • init(channel方法里面主要涉及到将Parent Channel 和 Child Channel的option和attribute 设值，并将客户端设置的参数覆盖到默认参数中；最后，还将childHandler(new TelnetServerInitializer())中设置的handler加入到pipeline()中。

  • channel.unsafe().register(regFuture) 把ServerBootstrapAcceptor这个Handler加入到Pipeline中

doBind0方法内部执行了javaChannel().register(eventLoop().selector, 0, this); 触发了服务端的channelActive() 事件，并设置了 selectionKey.interestOps(SelectionKey.OP_ACCEPT);

参考

Wiki Event loop

Architecture of a Highly Scalable NIO-Based Server

nio框架中的多个Selector结构

官网的相关文章

User Guide For 5.x

Core J2EE Patterns - Intercepting Filter

blog comments powered by Disqus